rwang220/Yida-Model-14B

Yida-Model-14B: A Specialized Chinese Medical LLM

Yida-Model-14B is a 14.77 billion parameter language model developed by rwang220, built upon the Qwen/Qwen3-14B architecture. This model is a merged full-weight checkpoint, resulting from LoRA supervised fine-tuning on an extensive internal Chinese medical instruction dataset. It is designed for research and engineering evaluation in the medical domain, specifically for tasks involving Chinese medical language.

Key Capabilities and Features

• Specialized Medical Domain: Fine-tuned on a diverse mix of Chinese medical tasks, including clinical, diagnostic, ethical, and record-keeping categories.
• Bilingual Support: Primarily focused on Chinese medical language but also supports English.
• Qwen3 Base: Inherits the robust architecture of Qwen3-14B, including its chat template for structured conversations.
• High Context Length: Supports a maximum context length of 40,960 tokens, with SFT training utilizing 8,192 tokens.
• Apache 2.0 License: Inherits the permissive Apache 2.0 license from its base model.

Intended Use Cases

• Chinese Medical Dialogue: Generating and understanding conversations in a medical context.
• Chart and Report Drafting: Assisting in the creation of medical charts and reports.
• NLP Prototypes: Developing natural language processing applications for the medical field.

Limitations

It is important to note that Yida-Model-14B is not a medical device. It may hallucinate information such as guidelines, doses, diagnoses, and citations. Users should not rely on it as the sole basis for clinical decisions. The training data is internal and not fully documented, and public clinical benchmarks are not claimed for this model.